Development of Embedded Machine Learning Finger Number Recognition Application using Edge Impulse Platform
Chun-Ki Kwon · 2023
This work demonstrates the use of a user-friendly AI platform-Edge Impulse-to develop and evaluate Tiny Machine Learning (TinyML) models. Edge Impulse is a platform for building TinyML models that are optimized to run efficiently on any tiny embedded device, and it therefore has tough resource constraints, such as a memory size of a few hundred kilobytes and ultra-low power consumption. To achieve our stated purpose, we focused on five finger number gestures representing the numbers from one to five, and we then used the Edge Impulse platform to build optimized machine learning applications to recognize these five selected finger number gestures. The Arduino Nano 33 BLE Sense is the target device on which a TinyML generated by an Edge Impulse platform is assumed to be running. In terms of performance, the TinyML was able to correctly recognize 92% of the finger number gestures from one to five.